nlcom for parameter estimates of different regressions

[Rose]

Hi,

I am running three separate regressions. There are variables that are the same in all three regressions. I am interested in doing some post-estimation checks using the nlcom command. How would I be able to call the regression coefficients from previous regressions to run this test? I understand from the help command that the _b[name] uses only the regression coefficient from the most recent regression, and I could specify the name of the regression if I use the sureg command, which I am not.

For example, I run model 1 and then model 3. In both regressions, I have the coefficient "education." How can I do a nlcom test from the education coefficient in regression 1 and the education coefficient in regression 3?

Thank you so much in advance!

[Maarten Buis]

The easiest solution is to estimate them all in one model. Often people ask such questions when they estimate different models on different sub-samples, e.g. men and women. In that case they can just as well estimate one model and include interaction effects.

[habib hussain khan]

Hi
I have the same issue. I have to estimate 3 different models all with xtabond2. I then want to compare the coefficients to see if there is a mediation effect. My first question is that can I run two model at the same time using xtabond2?? or is there any other way to check the mediation effect while using xtabond2?? I shall be really thankful. I am sorry Rose for I am posting my question in the this thread. It seemed relevant to my problem.